Wait the guy who wrote that tard was kidding yes?
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
PS4's memory subsystem has separate buses for CPU (20Gb/s) and GPU(176Gb/s)
- Thread starter alexandros
- Start date
Wait the guy who wrote that tard was kidding yes?
anexanhume
Member
It's 1.84 TF JESUS.
The few snippets I've seen in here so far are interesting so I'll check the whole article.
But the thread delivered already.
Regarding the download speed, I'm certain I've downloaded at my Internet max of 40mbps from PSN before. This test was on an "upto 20mbps" line and Steam only got 16mbps. I find it hard to believe they only had that line available to test on but whatever.
But the thread delivered already.
Regarding the download speed, I'm certain I've downloaded at my Internet max of 40mbps from PSN before. This test was on an "upto 20mbps" line and Steam only got 16mbps. I find it hard to believe they only had that line available to test on but whatever.
I went ahead and made a "clearer" version of the diagram posted on the first page.
Hope it helps.
Stop taking away our flops yo.
EDIT: Did you know that PS2 had about 6GFlops of computational power? Every Gflop counts, least of all dat xtra 23.
anexanhume
Member
Also, as I understand it, the purpose of Onion+ is to pass things like that constant data directly to the GPU from the CPU rather than having an intermediary memory layer in between them (GPU cache). No need for it to reside in GPU cache when the CPU cache is actually holding onto it.
I also assume the unified memory allows this to happen on a grander scale directly from the main memory when either needs to access the others' large chunk of data.
I also assume the unified memory allows this to happen on a grander scale directly from the main memory when either needs to access the others' large chunk of data.
Khalifa Jayy
Banned
" "The PS4's GPU is very programmable. There's a lot of power in there that we're just not using yet. So what we want to do are some PS4-specific things for our rendering but within reason - it's a cross-platform game so we can't do too much that's PS4-specific," he reveals.
"There are two things we want to look into: asynchronous compute where we can actually run compute jobs in parallel... We [also] have low-level access to the fragment-processing hardware which allows us to do some quite interesting things with anti-aliasing and a few other effects."
Fucking xbone holding back the ps4.
#unleashPS4 campaign, go!
"There are two things we want to look into: asynchronous compute where we can actually run compute jobs in parallel... We [also] have low-level access to the fragment-processing hardware which allows us to do some quite interesting things with anti-aliasing and a few other effects."
Fucking xbone holding back the ps4.
#unleashPS4 campaign, go!
That first pot! 1000 internet points for you, good sir.
Locuza
Member
I disagree with Cerny. I would call the PS4 AMD Station A and the Xbox One AMD Station B with some custom-stuff.
The FCL Enhancements, the unified memory access, the cache bypass, the eight ACEs and so on is more or less simply AMDs IP, without major changes from Sony.
Actually AMD calls it now FLC (Fusion Compute Link) and RMB (Radeon Memory Bus).
Maybe there was some influence from Sony, but the ACEs are not different from the one AMD is using in their products nowadays.
Kabini has 4 ACEs with 32 queues for 2 CUs
Bonaire/7790 has also 4 ACEs with 32 queues (the chart with 2 classic ACEs must be wrong)
I guess the Xbox One has also 4 ACEs with 32 queues.
And Kaveri will have like the PS4 8 ACEs with 64 queues for 8 CUs.
The real custom stuff about the GPU ist the second graphics command processor with high priority for the OS.
Not in the slightest... and sadly there are more posters like him - the only difference being they don't make their bs as obvious and sometimes hide it in walls of text...
phosphor112
Banned
The thing that puzzles me is the developers' mention of the need to allocate data correctly. When I first read about AMD's HSA, I assumed that everything would be unified and could be customized to the needs of the developers as required. I thought there wouldn't be any need at all for data separation, as to me and my limited knowledge it seemed that CPU and GPU data would be fed thrugh a common bus and allocated dynamically as needed. Now I'm confused
You can still do exactly that. HSA tries to optimize through generalization. One thing to keep in mind is that this design IS NOT the final HSA stage. The final HSA goal is to incorporate CPU and GPU into one... giant mixture of things so they can "know" what each other (CPU/GPU) are doing. These buses allow devs to do exactly that WITHOUT needing that final stage.
So, you have Garlic, which is just GPU to RAM. Perfect for graphics.
Then, more importantly, you have Onion and Onion+, that allows this "HSA" access to shine. You have a CPU process that needs to jump into compute or vice versa. Onion/+ allows the GPU to see what's in the CPU cache... and easily grab the data for use. No need to jump to the main RAM when it's already cached from the CPU ready to go to the GPU.
I rubbed butter and syrup all over dos GFLOPS and stuffed it in my belly.
BlastProcessing
Member
PS4: Chock full of Folic Acid
One massive advantage of unified memory is that the developers get to dictate how the RAM is used. With the PS3 the memory was set at a 50-50 ratio. With a unified system developers can tailor their allocations to their needs.
Hahah. I for one can't wait for those posters to start posting in this thread. This is going to be amusing.
It's good. Total 196GB bw. Though that bw can't compete against x1's infinite power of the cloud.
BruceLeeRoy
Banned
LOL!
IIRC correctly, on the Guerilla's slide about KZ SF, they said their code was only using one (1) CU...
Open_mouth
Insert Foot
OT: I honestly don't understand the most of these technical stuff but the ease of porting is awesome.
Go away with this adding of various bandwidths. There is no reason to get less precise.
It will just be used in fanboy wars about "See, the Sony people are advertising bullshit memory bandwidths too! LOL hypocritical!"
I believe as someone that is untechnical all you need to know is that John Carmack thinks that Sony "made wise engineering choices."
You're taking this way too seriously, benny.
TheGreyHulk
Member
there's nothing wrong with them having these separate buses.
you use the bus you need for the tasks at hand.
obviously the gpu is going to want to use the fat radeon bus for loading textures and all that good stuff.
some of the cpu stuff doesn't need a giant bus like the one used f/ the gpu. it would be pointless to have for the cpu alone.
this really doesn't change a thing for the ps4. this is NOT a bad thing
you use the bus you need for the tasks at hand.
obviously the gpu is going to want to use the fat radeon bus for loading textures and all that good stuff.
some of the cpu stuff doesn't need a giant bus like the one used f/ the gpu. it would be pointless to have for the cpu alone.
this really doesn't change a thing for the ps4. this is NOT a bad thing
TheRealTalker
Banned
I read a lot of these threads and I can just see the derails in the future provided by some well known posters that will allude to posts like the ones you make.
I'm just anticipating to be annoyed in the future, but I guess that is being too serious so you're correct.
fvng
Member
Not sure if that infinite power of the cloud is sarcasm or not (I suspect it is) but I remember Cerny calling bullshit on cloud computing being able to boost graphical fidelity?
http://www.youtube.com/watch?feature=player_detailpage&v=rCVplpC1YXY#t=157s
Locuza
Member
Depends on which IP-Level MS was aiming.
Trinity-Level or Kaveri-Level.
I would guess they share more or less the same tech.
crazygambit
Member
He isn't actually. It wouldn't matter what the power of the Xbone were much higher than the PS4. They still wouldn't be able to do much specific stuff to its architecture because they have to keep the game multiplatform.
I've always thought the comment of developing for the lowest common denominator to be a bit misguided. Even if both consoles had the same power, multiplatform games would still perform worse than exclusives because they can't take advantage of specific hardware features of either one. Which is why I've never been too fond of having more than 1 top of the line console. And before you say competition is always better for the consumer, realize than in many cases, especially involving economies of scale, it isn't.
Correct me if I'm wrong, but this doesn't mean that the total BW is 196GB/s.
The memory itself has a maximum bandwidth of 176GB/s.
The GPU has access to the full available bandwidth.
The CPU has access to the same pool of memory, but through a slower bus with a maximum of 20GB/s.
Right or wrong?
You are correct. It's also worth nothing that GCN GPUs are rated to be able to only handle 153GB/s so the division between the CPU and GPU is pretty much perfect for the bandwidth.
Ah, so, fewer conflicts over resources, gotcha.
I don't want to prohibit discussion, but - no offense - your thread's title and content due to your presumably lacking knowledge of the topic often leads / will lead (and or already led) to clueless people posting nonsense.
Fortunately there are competent posters who already clarified the matter, so no hard feelings.
phosphor112
Banned
the real question is if MS special sauce can be as tasty
OH GOD. LOOKS SO GOOD.
And I'm fasting. =[
gundamkyoukai
Member
They will take advantage of both systems just not at launch maybe in next 2 years you will start seeing systems advantages being played out .
LiquidMetal14
hide your water-based mammals
Yup both retail 7850 & 7870 operate with bandwidth of 153.6GB/s. The GPU in PS4 would at the least have access to an extra 2.4GB/sec at any given time. Plus, the CPU won't be constantly accessing data at that rate 100% of the time anyway.
Yeah, okay. I questioned myself about it since a lot of people thought the bandwidth to the GPU wouldn't be compromised.
Theoretically speaking, if the CPU utilizes the maximum bandwidth of 20GB/s, the GPU has access to a maximum of 156GB/s.
And all of this is routed through the 6x Blu-ray interface cutting it down to a measly 216MB/sec. That, it turn, is farmed out to the cloud at a rate of 1.5MB/sec, calculating in the average latency and work-time in the cloud, and it all effectively forms a minute black hole tearing apart the space-time continuum for each and every operation.
Sony am doomed.
Evolution_1ne
Banned
banning him was like getting rid of the float in mid air glitch in PS Home, just a chuck of fun gone forever
Ah man, that sucks. I heard they were aiming for one core for the OS.
I can understand xbone needing two cores for their 2 and a half OS setup but I'm perplexed as to why PS4 would need two as well.
TheRealTalker
Banned
OldAsUrSock
Banned
You are right. Cerny has called out the Cloud that it cannot increase graphical fidelity.
Dragon1893
Member
Do some people actually get pleasure out of doing fools out of themselves? I don't get it.
Anyway, interesting info that gives us a clearer picture of the PS4's hardware.
You know this wasn't anything special just confirming something that anyone with a little knowledge knew.
Sky is not purple
Can Crusher
Banned
Why do they need to reserve 2 cpu cores for the damn OS?
I am assuming for back ground stuff, multi-tasking and social sharing. It does not matter any way multiplatform be around 6 cores anyways. Also Sony first party will be using compute so the extra core not going to be that big of a deal. The extra ACE units easily make up for this.
Most probably because some 3rd party multimedia programs will be hosted there [Netflix].
Similar threads
- Locked
- Poll